Had a strange issue at a customer yesterday. The customer is running a set of clustered SMTP devices for inbound and outbound SMTP relay and we had added both IP addresses of these clustered devices to the relay server field on the Configuration document of both their clustered Domino servers. Yesterday one of the clustered SMTP devices went down but unfortunately Domino didn’t immediately switch to the other SMTP device. Doing a “tell router status” indicated that the router task was waiting for the device to come back up. As it was it looked as though it was unaware of the second entry for the SMTP relay server.
The Domino Administrator help file states under “Configuring multiple relay hosts”:
To enable greater control over outbound message routing, you can configure multiple relay hosts. Using multiple relay hosts enables Domino to route mail addressed to certain Internet domains to certain relay hosts, without first performing a DNS lookup. For example, you can split external SMTP mail routing so that Domino routes all outbound Internet mail along one path, except mail addressed to a specific domain, such as *.acmepartner.com, which it sends through a specific SMTP server. To configure multiple relay hosts, create a Foreign SMTP Domain document for each set of destinations, and then create SMTP connection documents to match these foreign SMTP domain documents. For example, using the previous example, you would create one Foreign SMTP Domain document for *.* and another for *acmepartner.com.
From the description I gather that Domino only can use one relay server and it you have multiple relay hosts you can only distribute the traffic among the servers based on domain name matches e.g. all .dk and .net traffic to one server and the rest to the other. But that’s load balancing and not failover.
Fortunately the customer have some intelligent network switches where you can define a logical interface (which is the one you would specify on the Configuration document) and have the switch distribute the traffic to the multiple devices based on which devices are up (the switch senses which of the devices respond to SMTP). Hence the switch does the failover and not Domino. This approach works but I was expecting Domino to provide this kind of functionality.
